AutoHire: Agentic AI-Powered Job Application Automation for Indian Tech

The structural gap between the supply of engineering graduates and the availability of domain-aligned employment in India motivates the need for intelligent, candidate-facing job application tooling. Existing research addresses the recruiter side of the hiring pipeline—resume parsing, candidate ranking, and HR automation—while the job-seeker is left without automated, quality-controlled application support. A further limitation of prior work on LLM-based career document generation is the absence of any mechanism to verify generated content against the candidate’s actual credentials, a gap that introduces factual inaccuracy into submitted materials. This paper presents AutoHire, a self-hosted, end-to-end agentic system designed to assist Indian tech freshers throughout the full job application lifecycle. AutoHire comprises four principal components: a structured resume intelligence pipeline with Pydantic-validated output; a five-dimensional weighted job scoring engine producing interpretable match scores; a two-pass anti-hallucination cover letter generator that verifies every generated claim against the source resume before submission; and a ReAct-based Planner–Actor–Validator browser automation agent with confidence-gated human escalation and field-level crash recovery. Experiments were conducted at the module level: resume parsing was evaluated on 200 resumes achieving a macro-averaged F1 of 0.84; job scoring was evaluated on 400 resume–job pairs achieving a Spearman correlation of 0.79; hallucination verification was evaluated on 150 generated letters reducing the hallucination rate from 34% to 3%; and browser automation was evaluated over 120 form-fill runs achieving an 83% end-to-end completion rate. The system natively targets Internshala, Wellfound, and Naukri—Indian portals not addressed by any prior system in the reviewed literature.